Creates a short narrative explaining a candidate's fit for a job.
| Time | Status | User Agent | |
|---|---|---|---|
Retrieving recent requests… | |||
Important: This version of
match-narrativealigns to Evaluate Candidates v5. If used with a different version ofevaluate-candidates, the narrative and match scores may not align.
Use this endpoint to create a Match Narrative, a natural language summary that helps users interpret an evaluation (e.g., provided by evaluate-candidates) of a candidate-job pair. Match Narratives help users to instantly understand a Candidate's fit for a Job, streamlining hiring and job searches.
Each narrative includes discrete sections on the candidate's work history, skills and education and with supporting citations of relevant work experience, important skills, and certifications that a candidate does or does not have, the relevance of a degree program, or other details supporting the evaluation.
AdeptID uses a Large Language Model (LLM) paired with AdeptID’s custom assessments to generate the narrative. AdeptID recommends that customers accompany the match narrative with a short statement that clearly and conspicuously states the narrative text was produced with generative AI.
Additional Notes:
- Enabling this service: This is a value-added service and is not available by default. To access this service, please contact your AdeptID Partner Success team.
- Intended Usage: The
match-narrativeendpoint is intended to be called asynchronously from a search or matching API call, to populate a more detailed UI on demand (instead of being called for every search or match result). This is a synchronous endpoint with a latency of multiple seconds; users should consider this when configuring API timeout limits. - Mitigating Bias: Instead of allowing an LLM to derive any evaluation of fit between candidate and job, an LLM is provided with quantitative evaluations from the audited AdeptID models; the LLM then converts the model outputs into a natural language format. This helps to avoid common hallucination and bias risks associated with LLMs.
- Privacy: The data provided to an LLM for Match Narrative generation is not utilized in the training of subsequent LLMs. For more information on AdeptID’s approach to data privacy and security, see https://trust.adept-id.com/.
- Languages: The output language is specified using the IETF BCP 47 (e.g., “en-US”) standard. See Languages for a list of acceptable codes.